8  Data sources

8.1 StreetLight Data

StreetLight Data is a transportation analytics platform that uses aggregated location-based services (LBS) data from cell phones and navigation/GPS data to deliver insights on travel patterns. For this project, we used StreetLight to find the volume of traffic (number of vehicles) and average trip length for passenger and commercial vehicles.

For ease of access, we used {streetlightR} to interact directly with the StreetLight API. {streetlightR} is an open-source R package maintained by Council staff.

Data source description, type

  • StreetLight Data, falls in second rank. Quality rank (See Table B.2)
  • How, when, and why was the data collected?
  • If this is a modeled dataset, what is the sample?
  • What is the raw unit of measurement?
  • How was this data accessed? Include any relevant links/citations, code, or downloads.
  • What data cleaning or wrangling was completed? How did you test these processes and outputs?
  • What is the geographic and temporal scope? Did you complete any aggregation?
  • What version is the data? Were there other versions available? If so, why did you choose this version?
  • What assumptions are made when we use this dataset?
  • Which subject matter expert (SME) reviewed this data?
  • Describe testing used to verify data

Be sure to add a citation of this dataset to the Zotero shared library.

8.1.1 Data characteristics

  • Were there any missing data? How did you handle missing data?
  • Plots, tables, and description of data distribution
  • Variance, Z-Score, quantiles
  • Facet views by categorical variables

8.1.2 Passenger

For passenger data, we used StreetLight Volume - an estimate of the number of vehicles. The models that make up StreetLight Volume predict vehicle volumes by combining Location-Based Services (LBS) trips with contextual features that represent the street network, demographics, climate, and other geographic characteristics (StreetLight Data 2023a). The models are validated against data permanent traffic counters across the country, including in the study area.

Table 8.1:

StreetLight passenger travel analysis sample size

Data periods Mode of travel Approximate device count Approximate trip count
Jan 01, 2021 - Dec 31, 2021 All Vehicles LBS Plus - StL All Vehicles Volume 1,038,000 125,411,000

8.1.2.1 Trip length validation

StreetLight returns not only vehicle volume, but also trip attributes, like trip length. We then use this to estimate vehicle miles traveled, by multiplying volume by average trip length for each origin-destination pair.

To test logical validity of average trip lengths, we will compare the minimum distance between each origin and destination with the average trip length. These should correlate.

In cases where the origin and destination counties are not adjacent, the average trip length is consistently higher than the minimum distance between the counties.

Figure 8.1: Avg. trip distance and minimum distance between counties
<<<<<<< HEAD

We can also compare these distances with the observed average trip distance from the Met Council Travel Behavior Inventory (TBI). Read more about the TBI in Section 8.5.

Figure 8.2: Avg. trip distance, Travel Behavior Inventory and StreetLight
=======
>>>>>>> origin/main

We would also expect that large counties will have longer trip lengths and smaller counties will have shorter trip lengths.

Comparing trip distance and county area, we see a general positive correlation (the larger the county, the longer the average trip).

<<<<<<< HEAD
Figure 8.3: Avg. distance for trips within county and county area
=======
Figure 8.2: Avg. distance for trips within county and county area
>>>>>>> origin/main

8.1.3 Commercial

StreetLight does not provide StreetLight Volume for 2021 commercial vehicle analyses. To measure volume for commercial traffic, we used the StreetLight Index, a relative measure of traffic volume, calibrated using AADT values to result in traffic volume (StreetLight Data 2023b).

StreetLight compares the AADT calibration values for a given zone with StreetLight’s sample size for the same zone, and creates a calibration factor to apply to the entire analysis (StreetLight Data 2023d). We generated a calibration zone set for commercial traffic by selecting road segments with both AADT and vehicle classification data in both MN and WI counties within the CPRG study area. Read more about state DOT vehicle weight distribution data in Section 8.3.0.2.

<<<<<<< HEAD
Table 8.2:

StreetLight commercial travel analysis sample size

Data periods Mode of travel Vehicle weight Approximate device count Approximate trip count
Jan 01, 2021 - Dec 31, 2021 Truck - StL Calibrated Truck Index Medium N/A 1,514,000
Jan 01, 2021 - Dec 31, 2021 Truck - StL Calibrated Truck Index Heavy N/A 605,000
=======

Then, we selected only the stations within the study area with observations in the last five years (2017-2021). Finally, we joined this data with Average Annual Daily Traffic (AADT) WisDOT (2021) road segments by station ID. The road segments sampled include multiple road functional classes and segments in all seven metro counties. All traffic sensor stations pulled were permanent, continuous monitoring sites.

Figure 8.3: StreetLight calibration locations and values
Figure 8.4: Vehicle weight distribution at calibration points
>>>>>>> origin/main

8.1.3.1 Calibration

StreetLight classifies commercial vehicles by Federal Highway Administration (FHWA) gross vehicle weight range (GWVR) classes: where vehicles between 14,000 lbs and 26,000 lbs (Class 4 to Class 6) are considered medium-duty, and vehicles greater than 26,000 lbs (Class 7+) are heavy-duty (StreetLight Data 2023c).

EPA’s Motor Vehicle Emissions Simulator (MOVES4) has their own, slightly different vehicle classification system (USEPA 2023b).

After reviewing MnDOT’s visual definitions of commercial vehicles, we defined MnDOT vehicle types 4-7 as medium-duty and types 8-13 as heavy-duty. We believe this configuration aligns most closely with both StreetLight’s and MOVES4’s vehicle classifications schemes.

However, vehicles falling in FHWA class 7 (> 26,000 lbs, < 33,000 lbs) are classified as medium duty by state DOTs, and heavy duty by StreetLight. This discrepancy is relatively small, and is unlikely to heavily influence emisssions reported.

<<<<<<< HEAD
=======
>>>>>>> origin/main
Table 8.3:

Vehicle weight classifications by data source

Gross vehicle weight rating (lbs) FHWA DOT StreetLight
<6000 Class 1: <6,000 lbs Light-duty Light-duty
10,000 Class 2: 6,001-10,000lbs Light-duty Light-duty
14,000 Class 3: 10,001-14,000 lbs Light-duty Light-duty
16,000 Class 4: 14,001-16,000 lbs Medium-duty Medium-duty
19,500 Class 5: 16,001-19,500 lbs Medium-duty Medium-duty
26,000 Class 6: 19,501-26,000 lbs Medium-duty Medium-duty
33,000 Class 7: 26,001-33,000 lbs Medium-duty Heavy-duty
>33,000 Class 8+: >33,001 lbs Heavy-duty Heavy-duty

To create the calibration dataset, we found the ratio of passenger/medium/heavy-duty vehicles for at traffic count stations within our study area using state DOT data. You can read more about vehicle classification data in Section 8.3.0.2.

8.2 Total vehicle volume validation

To validate our county origin-destination VMT data, we can compare the county totals to the DOT reported values from MnDOT (MnDOT 2021a) and WisDOT (WisDOT 2021). Note that these data include all vehicle types, both passenger and commercial.

Figure 8.4: County vehicle miles traveled and StreetLight Volume

8.2.1 Limitations

  • The data used for passenger traffic contains “all vehicles”, not just passenger vehicles, meaning that commercial vehicles may be double counted. As a default, StreetLight suggests that users use a ratio of 96/2/2 (95% passenger, 2% medium, 2% heavy). We could apply a scaling factor of 0.96 to the passenger data to account for this.
  • Commercial vehicle classifications schemes differ across data sources.

8.2.2 Comparison with similar datasets

8.3 State DOT data

As required by federal law, Minnesota and Wisconsin state departments of transportation (MnDOT and WisDOT) report various traffic measures for planning, forecasting, and various analysis endeavors.

8.3.0.1 Vehicle miles travled

Vehicle miles traveled (VMT) is a standardized measure created by multiplying average annual daily traffic (AADT) by centerline miles. AADT is an estimate of the total vehicles on a road segment on any given day of the year in all directions of travel. VMT and AADT are common traffic measures and standardized across the United States.

MnDOT and WisDOT derive VMT using traffic counts from continuous and short term traffic monitoring sites. These raw counts are adjusted by multiplying seasonal, day-of-week, and axle adjustment factors WisDOT (2023). Data is not collected for every site every year, but the data are sufficient for year-over-year comparisons.

These data were compiled from MnDOT and WisDOT county level reports. MnDOT provides Excel workbooks with VMT by county and route system on their website. These were downloaded, filtered to include on the relevant counties, and aggregated to the county level by summing VMT by county/route system. Processing code can be found in mndot_vmt_county.R.

WisDOT publishes PDF tables with county-level VMT. These were downloaded and data extracted using {tabulizer}, an R package interfacing with the Tabula PDF extractor library. Processing code can be found in wisdot_vmt_county.R.

Figure 8.5: County vehicle miles traveled
<<<<<<< HEAD
=======
>>>>>>> origin/main

Data source description, type

  • Quality rank (See Table B.2)
  • How, when, and why was the data collected?
  • If this is a modeled dataset, what is the sample?
  • What is the raw unit of measurement?
  • How was this data accessed? Include any relevant links/citations, code, or downloads.
  • What data cleaning or wrangling was completed? How did you test these processes and outputs?
  • What is the geographic and temporal scope? Did you complete any aggregation?
  • What version is the data? Were there other versions available? If so, why did you choose this version?
  • What assumptions are made when we use this dataset?
  • Which subject matter expert (SME) reviewed this data?
  • Describe testing used to verify data

Be sure to add a citation of this dataset to the Zotero shared library.

8.3.0.1.1 Data characteristics
  • Were there any missing data? How did you handle missing data?
  • Plots, tables, and description of data distribution
  • Variance, Z-Score, quantiles
  • Facet views by categorical variables
8.3.0.1.2 Limitations
  • Usually only samples county and state roads, primary arterials
  • Not every site is sampled every year

8.3.0.2 Vehicle distribution by weight

To calibrate the generalized StreetLight Index to get commercial vehicle counts, we created a set of spatial lines (roads) to calibrate StreetLight’s metrics. For each calibration road, we found the proportion of passenger, medium-, and heavy-duty vehicles in the most recent available year, up to 2021.

State DOTs operate vehicle classification stations, which provide both the volume of traffic on a given road segment and, for some locations, the breakdown of volume by vehicle type. We obtained this breakdown using data from MnDOT (MnDOT 2021b) and WisDOT (WisDOT 2020) reporting websites.

MnDOT provides AADT road segments, which align with station identification numbers. Wisconsin does not readily supply AADT road segment data - as suggested by the Wisconsin cartographers office (State Cartographer’s Office 2021), we pulled OpenStreetMaps road data (OSM version 0.6).

Then, we selected only the stations within the study area with observations in the last five years (2017-2021). Finally, we joined this data with Average Annual Daily Traffic (AADT) WisDOT (2021) road segments by station ID. The road segments sampled include multiple road functional classes and segments in all counties. All traffic sensor stations pulled were permanent, continuous monitoring sites.

Figure 8.6: StreetLight calibration locations and values
Figure 8.7: Vehicle weight distribution at calibration points

Introduction text Data source description, type

  • Quality rank (See Table B.2)
  • How, when, and why was the data collected?
  • If this is a modeled dataset, what is the sample?
  • What is the raw unit of measurement?
  • How was this data accessed? Include any relevant links/citations, code, or downloads.
  • What data cleaning or wrangling was completed? How did you test these processes and outputs?
  • What is the geographic and temporal scope? Did you complete any aggregation?
  • What version is the data? Were there other versions available? If so, why did you choose this version?
  • What assumptions are made when we use this dataset?
  • Which subject matter expert (SME) reviewed this data?
  • Describe testing used to verify data

Be sure to add a citation of this dataset to the Zotero shared library.

8.3.0.2.1 Data characteristics
  • Were there any missing data? How did you handle missing data?
  • Plots, tables, and description of data distribution
  • Variance, Z-Score, quantiles
  • Facet views by categorical variables
8.3.0.2.2 Limitations

8.3.1 Data dictionaries

<<<<<<< HEAD
=======
>>>>>>> origin/main
Table 8.4:

County vehicle miles traveled metadata

Column Class Description
year character VMT estimation year
county character County name
daily_vmt numeric Vehicle miles traveled on an average day
annual_vmt numeric Annual vehicle miles traveled
centerline miles numeric Centerline miles included in VMT estimate. Minnesota only
data_source character State DOT. Either "MnDOT" or "WisDOT"

Table with detailed description of columns and definitions for each data table.

8.4 EPA MOVES

Emissions rates for our region were calculated using the EPA’s Motor Vehicle Emissions Simulator (MOVES4) (USEPA 2023a). MOVES4 calculates emissions factors using Council’s regional travel demand model, Minnesota Department of Vehicle Services county vehicle registration data, and the Minnesota Pollution Control Agency vehicle age distribution. Each of these inputs helps the model estimate the characteristics of vehicles on the road in our region. The model takes into account differences in fuel economy (miles per gallon) depending on a vehicle’s age and size, as well as its fuel intake (diesel or gasoline). The results are specific to the conditions of our region, and so are more accurate than national averages.

CO2 equivalence (CO2e) values are derived from global warming potential (GWP) values. See Section A.2 for more details.

<<<<<<< HEAD
=======
>>>>>>> origin/main
Table 8.5:

Emissions in grams per vehicle mile traveled by vehicle weight in the Twin Cities region. EPA MOVES4

Vehicle Weight MOVES Year Grams CH4 per mile Grams N2O per mile Grams CO2 per mile Grams CO2e per mile
Passenger 2019 0.0100 0.0100 353.43 356.36
Medium 2018 0.0100 0.0100 473.24 476.17
Heavy 2018 0.0206 0.0402 1212.36 1223.59
<<<<<<< HEAD
=======
>>>>>>> origin/main
Table 8.6:

Grams of emissions per mile by vehicle type and fuel type. EPA GHG Emission Hub (2021)

Vehicle Type Vehicle Year Fuel Type Grams CH4 per mile Grams N2O per mile Grams CO2 per mile Grams CO2e per mile
Passenger Car 2013 Gasoline 0.007 0.005 364.315 365.733
Passenger Car 2014 Diesel 0.030 0.019 315.123 321.057
Light Truck 2007 Diesel 0.001 0.001 461.991 462.387
Heavy-Duty Vehicle 2007 Diesel 0.009 0.043 787.809 799.496

Data source description, type

  • Quality rank (See Table B.2)
  • How, when, and why was the data collected?
  • If this is a modeled dataset, what is the sample?
  • What is the raw unit of measurement?
  • How was this data accessed? Include any relevant links/citations, code, or downloads.
  • What data cleaning or wrangling was completed? How did you test these processes and outputs?
  • What is the geographic and temporal scope? Did you complete any aggregation?
  • What version is the data? Were there other versions available? If so, why did you choose this version?
  • What assumptions are made when we use this dataset?
  • Which subject matter expert (SME) reviewed this data?
  • Describe testing used to verify data

Be sure to add a citation of this dataset to the Zotero shared library.

8.4.1 Data characteristics

  • Were there any missing data? How did you handle missing data?
  • Plots, tables, and description of data distribution
  • Variance, Z-Score, quantiles
  • Facet views by categorical variables

8.4.2 Limitations

8.4.3 Comparison with similar datasets

8.4.4 Data dictionary

<<<<<<< HEAD
=======
>>>>>>> origin/main
Table 8.7:

EPA MOVES metadata

Column Class Description

moves_year

numeric

EPA MOVES model run year

vehicle_weight

c(“ordered”, “factor”)

Vehicle weight classification

co2

numeric

Grams of carbon dioxide emitted per vehicle mile traveled

co2_co2_equivalent

numeric

Grams of carbon dioxide and carbon dioxide equivalent according to global warming potential (GWP)

ch4

numeric

Grams of methane emitted per vehicle mile traveled

n2o

numeric

Grams of nitrous oxide per vehicle mile traveled

Table with detailed description of columns and definitions for each data table.

8.5 Travel Behavior Inventory

The Metropolitan Council Travel Behavior Inventory (TBI) is a bi-annual household survey around 7,500 families in the 7-county Twin Cities metro and three neighboring Wisconsin counties. Information on people, households, trips, and vehicles are collected.

8.5.1 Regional fleet characteristics

We used 2021 TBI data to determine the regional average vehicle age and distribution of diesel and gasoline passenger vehicles.

Vehicles were classified into broad fuel categories - diesel and gas + all other fuels (including gasoline, electric, flex-fuel, hybrid, and plug-in hybrid) to best match the average miles per gallon table specifications in the EPA Local Greenhouse Gas Inventory Tool. The resulting value is on par with recent statistics from the Bureau of Transportation Statistics (BTS), which calculates the average passenger vehicle age in 2021 to be 12.1 years (BTS 2023).

TBI data were cleaned to only include vehicles with complete data and model year 1980 or later.

<<<<<<< HEAD
=======
>>>>>>> origin/main
Table 8.8:

Median vehicle age and proportion of all regional vehicles by fuel type

Fuel type Median vehicle year Median vehicle year standard error Estimated number of vehicles Estimated number of vehicles standard error Estimated percentage of all vehicles Estimated percentage of all vehicles standard error Sample size
Gas + all other fuels 2013 0 2,256,249 60,371 98.78% 0.21% 10,451
Diesel 2014 2 27,947 4,918 1.22% 0.21% 122
Figure 8.8: Regional vehicle fleet model year by fuel type

8.5.2 Average trip distance between counties

The average trip distance for the entire region is 6.27 miles (standard error 0.14), based on a sample of 117,568 trips.

Table 8.9:

Mean trip distance by origin-destination county

Origin-Destination pair Sample size Mean trip distance (miles) Mean trip distance standard error Estimated number of vehicles Estimated number of vehicles standard error
Anoka-Anoka 4,476 3.52 0.20 613,754 30,689
Anoka-Carver 5 32.81 5.18 100 63
Anoka-Chisago 66 27.29 4.86 11,404 3,744
Anoka-Dakota 61 27.49 2.28 3,814 1,669
Anoka-Hennepin 1,261 11.97 0.65 149,460 14,876
Anoka-Ramsey 717 9.89 0.73 81,497 12,577
Anoka-Scott 25 33.48 4.44 1,527 627
Anoka-Sherburne 104 14.51 1.75 22,934 6,720
Anoka-Washington 108 22.20 2.31 11,324 2,820
Carver-Anoka 6 33.28 4.39 131 76
Carver-Carver 2,098 8.57 3.73 144,426 14,513
Carver-Chisago 2 NaN NaN 0 0
Carver-Dakota 48 34.51 0.83 4,367 2,183
Carver-Hennepin 610 12.29 1.01 72,092 9,621
Carver-Ramsey 17 31.04 0.38 1,412 1,152
Carver-Scott 110 6.79 1.06 13,873 3,750
Carver-Washington 16 39.31 0.40 1,431 562
Chisago-Anoka 60 17.61 3.94 5,806 2,519
Chisago-Carver 2 NaN NaN 0 0
Chisago-Chisago 510 7.82 1.77 49,698 7,362
Chisago-Dakota 15 46.62 2.71 156 84
Chisago-Hennepin 48 49.92 3.57 2,886 1,696
Chisago-Ramsey 57 33.40 0.83 2,921 1,632
Chisago-Sherburne 1 NaN NaN 0 0
Chisago-Washington 144 12.88 2.13 14,555 3,566
Dakota-Anoka 51 25.23 0.89 3,557 1,237
Dakota-Carver 54 29.35 0.64 5,160 2,602
Dakota-Chisago 16 45.18 1.21 261 113
Dakota-Dakota 9,427 3.22 0.11 792,192 30,615
Dakota-Hennepin 1,335 14.30 0.64 87,100 8,382
Dakota-Ramsey 1,124 9.48 0.64 71,539 8,617
Dakota-Scott 341 8.95 1.01 43,366 8,317
Dakota-Sherburne 7 83.05 0.00 519 519
Dakota-Washington 358 15.19 1.75 29,759 6,024
Hennepin-Anoka 1,279 12.11 0.80 149,876 15,672
Hennepin-Carver 593 12.04 1.16 72,544 9,776
Hennepin-Chisago 56 53.25 2.73 1,868 1,422
Hennepin-Dakota 1,352 14.66 0.69 97,537 9,735
Hennepin-Hennepin 46,461 4.26 0.07 2,984,065 48,069
Hennepin-Ramsey 3,985 10.87 0.56 178,613 10,220
Hennepin-Scott 394 10.76 1.16 47,265 7,937
Hennepin-Sherburne 134 27.64 1.68 12,018 2,491
Hennepin-Washington 296 20.31 1.18 23,747 4,375
Ramsey-Anoka 702 10.58 0.70 74,670 12,067
Ramsey-Carver 22 35.72 0.97 1,606 1,165
Ramsey-Chisago 61 40.56 2.48 8,293 3,062
Ramsey-Dakota 1,120 9.99 0.64 76,215 9,282
Ramsey-Hennepin 4,027 11.20 1.31 183,555 10,688
Ramsey-Ramsey 20,532 3.36 0.13 1,097,772 27,449
Ramsey-Scott 58 29.36 2.66 1,911 1,073
Ramsey-Sherburne 26 46.18 2.35 1,714 676
Ramsey-Washington 1,438 11.76 1.38 100,203 8,667
Scott-Anoka 18 37.54 1.26 1,217 514
Scott-Carver 116 8.17 1.20 15,669 3,917
Scott-Chisago 1 75.23 0.00 1 1
Scott-Dakota 326 7.88 0.78 39,340 7,775
Scott-Hennepin 413 12.95 1.54 49,451 8,581
Scott-Ramsey 59 30.16 3.35 6,423 2,764
Scott-Scott 1,899 4.67 0.46 248,429 18,964
Scott-Sherburne 7 42.03 2.01 209 123
Scott-Washington 6 39.97 0.21 391 385
Sherburne-Anoka 89 14.21 2.64 8,554 2,896
Sherburne-Chisago 1 NaN NaN 0 0
Sherburne-Dakota 6 82.54 0.00 519 519
Sherburne-Hennepin 128 28.05 1.87 8,341 1,848
Sherburne-Ramsey 33 40.05 3.31 2,886 905
Sherburne-Scott 5 44.92 0.00 72 51
Sherburne-Sherburne 702 23.63 9.00 93,336 10,987
Sherburne-Washington 4 46.70 0.98 24 18
Washington-Anoka 107 21.55 2.19 12,164 2,730
Washington-Carver 16 59.59 0.40 2,554 1,815
Washington-Chisago 147 11.11 1.33 13,417 3,462
Washington-Dakota 348 15.85 1.42 31,206 6,520
Washington-Hennepin 325 21.17 1.26 25,356 4,282
Washington-Ramsey 1,411 9.25 0.54 94,126 8,216
Washington-Scott 4 41.92 0.00 277 277
Washington-Sherburne 3 46.20 0.50 24 18
Washington-Washington 5,578 3.45 0.12 521,261 21,560

Trips with distance over 720 miles (the equivalent of 12 hours of driving at 60 miles per hour) were removed.

Introduction text Data source description, type

  • Quality rank (See Table B.2)
  • How, when, and why was the data collected?
  • If this is a modeled dataset, what is the sample?
  • What is the raw unit of measurement?
  • How was this data accessed? Include any relevant links/citations, code, or downloads.
  • What data cleaning or wrangling was completed? How did you test these processes and outputs?
  • What is the geographic and temporal scope? Did you complete any aggregation?
  • What version is the data? Were there other versions available? If so, why did you choose this version?
  • What assumptions are made when we use this dataset?
  • Which subject matter expert (SME) reviewed this data?
  • Describe testing used to verify data

Be sure to add a citation of this dataset to the Zotero shared library.

8.5.3 Data characteristics

  • Were there any missing data? How did you handle missing data?
  • Plots, tables, and description of data distribution
  • Variance, Z-Score, quantiles
  • Facet views by categorical variables

8.5.4 Limitations

8.5.5 Comparison with similar datasets

8.5.6 Data dictionary

Table with detailed description of columns and definitions for each data table.